Ticket Terminal

Writing back a reviewed digest instead of the raw terminal log is the right default; the ticket stays readable. Is Notion usage anywhere near Jira so far? Search note: I googled "open jira ticket in claude code" (Google, US settings). Atlassian's own support page "Open a Jira work item in an AI coding tool" is #1, and their "Introducing Claude Agent for Jira" post is on page 1 too, alongside setup guides (adamgradzki.com, Swytchcode, MindStudio), an Atlassian community thread, a Reddit post and an X post. Ticket Terminal isn't there yet, and with Atlassian now covering the basic flow, the local memory graph is the part I'd make a page about. I run AI Knows Us, so search checks like this are what I do.

@Kris The raw transcript is not written back automatically. Ticket Terminal pulls the current Jira ticket, enriches it locally with the memory graph and relevant memories from previous tickets, and keeps the full transcript and agent trail attached to the ticket locally. For Jira, the manager’s digest can be reviewed and written back, so the ticket gets the useful outcome without dumping a noisy terminal log into it. Would you want that write-back to happen automatically or always require approval?

Keeping Jira as the source of truth while the agent works locally is the right call — copy-pasting ticket context into the terminal gets old fast. When the session finishes, does the transcript get written back to the ticket automatically?

I built Ticket Terminal because the ticket and terminal history kept becoming two separate stories. It keeps Jira or Notion as the source of truth, then opens the ticket as a real local Claude Code or Codex session with the transcript and agent trail attached. This is a public preview. It is single-user, setup still needs an authenticated CLI and tracker credentials, and it is not a sandbox. I’m mainly looking for people willing to try one real ticket and tell me where the setup or workflow breaks.